Rosie
The Product Innovation team sits on a deep library of end-user research — and the hard part was never collecting it, but synthesizing it on demand. Rosie was built to do exactly that: a deterministic research orchestrator that returns a grounded, four-lens read of MillerKnoll’s own research. Its POC then revealed something bigger about the data underneath it.
What it delivers
The POC did its real job: it proved the data was ready.
Rosie grew out of the Innovation Insights Experiment. Building it against the End User Research library did more than ship an agent — it revealed that the underlying research corpus was structured and rich enough to power a knowledge graph. On the strength of that finding, ongoing ownership moved from AISE to the Data Science team (Steve Meadows) to take the data in that direction. The AISE build stands in TEST; the next chapter is a data-science one. A POC that ends by pointing to a bigger opportunity is a POC that worked.
The principles
Deterministic over clever
Rosie routes with explicit, button-based selection rather than free-form intent guessing — so the path a query takes is predictable, testable, and repeatable across every researcher who uses it.
Grounded, structured, cited
The Internal Research capability answers from MillerKnoll’s own End User Research library across four fixed lenses, with evidence-rigor levels and required citations — synthesis you can trace, not a summary you have to trust.
Knowing when to hand off
The most valuable thing the POC produced was a finding about the data itself. Rather than over-build, AISE handed Rosie to Data Science to pursue the knowledge-graph opportunity it surfaced — the right team for the next problem.
Where it stands
Rosie is in the Copilot Agents TEST environment on Copilot Studio’s new agent experience (replatform completed June 26, 2026), with Jen Mackall as business owner in Product Innovation. It grew out of the Innovation Insights Experiment (LEANAI-38, closed). Ongoing technical ownership has transitioned to Steve Meadows on the Data Science team after the POC demonstrated the End User Research data was ripe for a knowledge graph; the broader productionization epic (ASE-389) is on hold pending that data-science direction.